Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 12 Nov 2004 00:13:53 -0700
From:      Scott Long <scottl@freebsd.org>
To:        John Baldwin <jhb@freebsd.org>
Cc:        gallatin@freebsd.org
Subject:   Re: New ACPI PCI Link Routing code
Message-ID:  <419462B1.6090103@freebsd.org>
In-Reply-To: <200411111737.00537.jhb@FreeBSD.org>
References:  <200411111737.00537.jhb@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help
John Baldwin wrote:
> I've done some work on the ACPI PCI link code to make it a bit more like $PIR 
> in that it is link centric and uses actual new-bus devices for each device 
> link.  One benefit of this is that unused links will be disabled now which 
> might help with interrupt aliasing problems on machines using APICs.  Also, 
> instead of routing IRQs for links via PCI device numbers using tunables, they 
> are now routed via the link name ala $PIR.  Thus, one uses 
> 'hw.pci.link.LNKA.irq=X' to route LNKA to IRQ X.  Also, when choosing a 
> virgin interrupt, we no longer try to guess at which IRQs might be used by 
> ISA devices.  Instead, we only use known-good IRQs including IRQs that the 
> BIOS has already used and the SCI if the link is being routed via ISA IRQs.  
> The patch is at http://www.FreeBSD.org/~jhb/patches/acpi_pci_link.patch  
> Please test and let me know if there are any problems, thanks.
> 

I had to hack up acpi_pcib_acpi.c in order to get it to compile.  The
result boots, but it doesn't seem to change anything and doesn't help
the interrupt aliasing issues that I have.  The dmesg is at
http://people.freebsd.org/~scottl/dmesg.wv1u.nov12.  The interrupt
aliasing manifests itself as irq 16 getting all of the interrupts from
both irq 30 and 49.  30 and 49 still get their interrupts, and they
don't get each other's interrupts.

Scott



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?419462B1.6090103>